Havells One now lets users control connected home appliances with natural speech in English and seven Indian languages. Havells India launched the Agentic AI Voice Mode on September 4, 2026, using ElevenLabs speech models as the voice layer while Havells’ own software interprets requests and sends commands to registered devices.

Key takeaways

  • Havells One supports English, Hindi, Tamil, Telugu, Bengali, Kannada, Gujarati and Malayalam in the new voice mode.
  • The feature is available to the app’s stated base of 2.67 million users, but it controls only compatible smart products registered in the app.
  • It can interpret natural or code-mixed speech, keep conversational context and coordinate more than one appliance.
  • Havells has not published independent accuracy, latency, privacy or device-compatibility test results.

What Havells One launched

Item Confirmed detail
Product Agentic AI Voice Mode in the Havells One app
Languages English plus seven Indian languages
Reported reach 2.67 million Havells One users
Voice technology ElevenLabs streaming speech-to-text and text-to-speech
Device logic Havells conversational intelligence and device-control agents

The supported products include connected fans, air purifiers, water purifiers and water heaters, Lloyd air conditioners and washing machines, and Crabtree smart switches. Havells says a user only needs a compatible smart device registered in Havells One to access the feature.

How a Havells One voice request reaches connected appliances A four-step flow from multilingual speech through recognition and Havells device-control agents to compatible registered appliances. Natural speech8 languagescode-mixing ElevenLabsspeech to texttext to speech Havells agentsintent + contextdevice commands Registeredsmartdevices

How Havells One differs from fixed commands

Traditional smart-home voice controls often depend on a precise device name and a fixed phrase. Havells says the new mode instead identifies intent, remembers context across turns and asks for clarification when a request is ambiguous. A broad request such as making a room comfortable can therefore be translated into actions across an air conditioner, fan, purifier and lights.

Everyone else is reporting the launch; we are explaining the control chain and its limits. The important change is orchestration: speech is converted into text, Havells’ agents decide which registered devices are relevant, and commands are sent across different connectivity protocols. That is closer to a household automation layer than a simple microphone button.

The language design is also practical for India. People routinely mix English with regional languages inside one sentence. The system is meant to handle that pattern rather than forcing a household to choose one language for every command. Related Lapaas Voice coverage of Alexa+ testing Hindi in India and Jio TeleFrame’s voice-first home strategy shows why local-language control is becoming a competitive layer.

What is still unproven

Havells says it evaluated recognition accuracy, multilingual support, latency and voice quality before moving to production in roughly six weeks. Neither Havells nor ElevenLabs published the test set, error rates, response times or results by language, so the announcement does not establish how reliably the system performs in noisy rooms, with different accents or on weak connections.

The companies also did not publish a detailed data-retention or voice-processing explanation in the launch material. Users should review current Havells One permissions and privacy disclosures before enabling voice features, especially in shared households. The launch confirms availability and intended capabilities, not independent proof of every claimed interaction.

Device control also creates a different safety standard from a conversational answer. A misunderstood request that changes a fan speed is inconvenient; one that operates heating, water or security-related equipment may carry greater consequences. Havells will need clear confirmations for sensitive actions, household permission controls and an accessible activity history so users can see which request triggered each device change.

What comes next

Havells plans to add more languages, compatible appliances, household-level personalisation, troubleshooting, scheduling and product FAQs. The strongest evidence will be a clear compatibility list, public privacy documentation and repeatable performance results across languages and real homes.

The concise answer is this: Havells One is using multilingual voice as a control layer for compatible connected appliances, allowing one natural-language request to trigger several device actions. Its value will depend on recognition accuracy, safe permissions and whether the automation behaves predictably outside a demonstration.

Frequently asked questions

Which languages does Havells One support?

The new mode supports English, Hindi, Tamil, Telugu, Bengali, Kannada, Gujarati and Malayalam. Havells says more languages are planned.

Does Havells One control every Havells appliance?

No. The voice mode works with compatible smart products registered in the app. Conventional, unconnected appliances are not made controllable merely by installing the app.

Who provides the voice AI?

ElevenLabs supplies streaming speech recognition and speech synthesis. Havells supplies the conversational intelligence and device-control agents that interpret intent and act on registered appliances.

Sources

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.